An automated approach for real-time informative frames classification in laryngeal endoscopy using deep learning.

PURPOSE: Informative image selection in laryngoscopy has the potential for improving automatic data extraction alone, for selective data storage and a faster review process, or in combination with other artificial intelligence (AI) detection or diagnosis models. This paper aims to demonstrate the feasibility of AI in providing automatic informative laryngoscopy frame selection also capable of working in real-time providing visual feedback to guide the otolaryngologist during the examination. METHODS: Several deep learning models were trained and tested on an internal dataset (n = 5147 images) and then tested on an external test set (n = 646 images) composed of both white light and narrow band images. Four videos were used to assess the real-time performance of the best-performing model. RESULTS: ResNet-50, pre-trained with the pretext strategy, reached a precision = 95% vs. 97%, recall = 97% vs, 89%, and the F1-score = 96% vs. 93% on the internal and external test set respectively (p = 0.062). The four testing videos are provided in the supplemental materials. CONCLUSION: The deep learning model demonstrated excellent performance in identifying diagnostically relevant frames within laryngoscopic videos. With its solid accuracy and real-time capabilities, the system is promising for its development in a clinical setting, either autonomously for objective quality control or in conjunction with other algorithms within a comprehensive AI toolset aimed at enhancing tumor detection and diagnosis.

Intra and interobserver agreement of narrow band imaging for the detection of head and neck tumors.

BACKGROUND: Narrow band imaging (NBI) in combination with white light endoscopy (WLE) has improved the accuracy for the diagnosis and follow-up of head and neck carcinomas by identifying changes in the vascular patterns of the mucosa. However, NBI evaluation is explorer-dependent and may be influenced by the learning curve. The aim of this study is to assess the intra and interobserver agreement of NBI and WLE at the office, under local anaesthesia, by either experienced or non-experienced observers. METHODS: Eighty-seven images of head and neck lesions were routinely collected under WLE and NBI. A group of three experienced otolaryngologists and three medical students assessed the images after a brief training. No additional patient information was provided. The same protocol was repeated after three weeks. Intra and interobserver agreement were calculated with the kappa index. RESULTS: NBI intraobserver agreement was substantial (κ = 0.62) and better than with WLE alone, which was moderate (κ = 0.57) in both groups. Interobserver agreement was moderate with WLE (κ = 0.58) and substantial with NBI (κ = 0.63). Both groups improved intraobserver and interobserver agreement with the implementation of NBI. CONCLUSIONS: Intra and interobserver agreement with NBI for the evaluation of head and neck lesions are substantial, and improve the results of WLE alone in both, professionals and trainees.

Timing, Complications, and Safety of Tracheotomy in Critically Ill Patients With COVID-19.

IMPORTANCE: The current coronavirus disease 2019 (COVID-19) pandemic has led to unprecedented needs for invasive ventilation, with 10% to 15% of intubated patients subsequently requiring tracheotomy. OBJECTIVE: To assess the complications, safety, and timing of tracheotomy performed for critically ill patients with COVID-19. DESIGN, SETTING, AND PARTICIPANTS: This prospective cohort study assessed consecutive patients admitted to the intensive care unit (ICU) who had COVID-19 that required tracheotomy. Patients were recruited from March 16 to April 10, 2020, at a tertiary referral center. EXPOSURES: A surgical tracheotomy was performed for all patients following recommended criteria for use of personal protective equipment (PPE). MAIN OUTCOMES AND MEASURES: The number of subthyroid operations, the tracheal entrance protocol, and use of PPE. Infections among the surgeons were monitored weekly by reverse-transcriptase polymerase chain reaction of nasopharyngeal swab samples. Short-term complications, weaning, and the association of timing of tracheotomy (early [≤10 days] vs late [>10 days]) with total required days of invasive ventilation were assessed. RESULTS: A total of 50 patients (mean [SD] age, 63.8 [9.2] years; 33 [66%] male) participated in the study. All tracheotomies were performed at the bedside. The median time from intubation to tracheotomy was 9 days (interquartile range, 2-24 days). A subthyroid approach was completed for 46 patients (92%), and the tracheal protocol was adequately achieved for 40 patients (80%). Adequate PPE was used, with no infection among surgeons identified 4 weeks after the last tracheotomy. Postoperative complications were rare, with minor bleeding (in 6 patients [12%]) being the most common complication. The successful weaning rate was higher in the early tracheotomy group than in the late tracheotomy group (adjusted hazard ratio, 2.55; 95% CI, 0.96-6.75), but the difference was not statistically significant. There was less time of invasive mechanical ventilatory support with early tracheotomy compared with late tracheotomy (mean [SD], 18 [5.4] vs 22.3 [5.7] days). The reduction of invasive ventilatory support was achieved at the expense of the pretracheotomy period. CONCLUSIONS AND RELEVANCE: In this cohort study, with the use of a standardized protocol aimed at minimizing COVID-19 risks, bedside open tracheotomy was a safe procedure for patients and surgeons, with minimal complications. Timing of tracheotomy may be important in reducing time of invasive mechanical ventilation, with potential implications to intensive care unit availability during the COVID-19 pandemic.